## A Tipsy Earth?

Hi. I’m Jonah, a scientist here at the UoR, and I study whether global warming will be more like drinking water, soda, or beer.

What do I mean by that? Let me explain.

Both thirst and the Earth’s climate are feedback systems – they react to an external nudge by either undoing or amplifying that initial push. A reaction that undoes the push is called a negative feedback, and an amplifying one a positive feedback.

For example, we all have experience being thirsty – in fact, you might be thirsty right now. How might you react? Perhaps you reach for a cold glass of water (Fig. 1). Drinking the water quenches your thirst. The initial nudge – your thirstiness – prompted you to react by doing something that got rid of your thirstiness; or put another way, the quenching nature of water provided a negative feedback.

Figure 1:  A glass of water: an absolute classic beverage.

But what if, instead, you reached for something sweet, like soda, or juice? Sweet drinks taste good, so that while they quench your thirst, they also stimulate you, making you want to drink more. The sweetness adds a positive feedback. The combined result is that drinking a sweet drink lessens your thirst, but not as much as just drinking water; the overall negative feedback is weaker. So you might find yourself drinking more soda than you would water. Fig. 2 shows a drink called a “Big Gulp” which is popular back in my home of the USA and contains about a litre of soda. Big Gulps of water are few and far between.

Figure 2: A Big Gulp: finally, enough soda.

Ah, but what if instead of a sweet drink, you decided to go for something with a bit of a kick? Some beer, cider, or whisky, for example (Fig. 3)? Depending on the drink, it may be more or less tasty – at first! But after a few drinks, the idea of getting another may become more appealing. In other words, with alcohol, the “tastiness” of the substance can depend on how much of it you’ve already consumed. What might start as a simple thirst can run away into a  bender, as your overall feedback goes from negative to positive.

Figure 3: Beer: another round, please!

What does any of this have to do with global warming? First, a bit of science. You might have heard that energy is conserved – it can be moved around, or change forms, but is never created or destroyed. One of the ways energy moves around is as light, which carries energy from the place it is radiated to the place it is absorbed. You may be used to light as something you can see, but there are actually many colours of light beyond the range our eyes can capture. Everything is constantly giving off light, invisible and/or visible, including you and me (doesn’t that give you a bit of glow?). And I’m not talking about reflecting visible light – you give off your own light, even in the dark, which is how some snakes are able to find you in the night (!).

The hotter you are, the more light you give off – and so the more energy you radiate away. The Sun is so hot (and huge) that even though it is quite far away, a substantial amount of the light it gives off reaches us here on Earth, where we gain the energy it carries. As a result, the Earth warms up – and gives off light of its own. The additional carbon dioxide we have been adding to the air for the past century or two intercepts some of the Earth’s light that was destined for space. As a result, the Earth doesn’t lose some of the energy it would otherwise have lost. This is kind of like us getting thirsty – it is a nudge, a push to a system that was previously in balance.

How does the Earth react? That excess energy warms the Earth’s surface, and as a result, it gives off more light, offsetting the carbon dioxide’s effect. In other words, we have a negative feedback, which we call the “Planck” feedback. This warming continues until the amount of light leaving the Earth once more balances the amount entering, and the planet is no longer thirsty. If the Planck feedback was the only one in the Earth’s climate, global warming would be less of an issue, because we’d only need to warm a little to restore balance (blue line in Fig. 4).

But like with soda, there’s a sweetness to warming that keeps the Earth going further: warming also causes more water vapor to be evaporated into the atmosphere. Water vapor also intercepts some of the Earth’s light destined for space, so it causes still more energy to stick around as excess. In other words, it is a positive feedback. It is not so strong as the Planck feedback, so that warming still manages to offset the initial carbon dioxide nudge; but instead of a little warming, we get more of a Big Gulp (red line in Fig. 4). And in addition to the water vapor feedback, there are others, positive and negative, associated with changes to ice, clouds, and the vertical structure of the atmosphere.

But here’s the twist: the water vapor feedback gets stronger the more you warm. In other words, it’s more like beer than soda. And the result is that too much of a push can send your planet on a bender. That’s what happened to Venus in the distant past, when the Sun’s increasing warmth caused its climate to enter a “runaway greenhouse,” evaporating its oceans and giving it a surface temperature of around 500ºC (gray line in Fig. 4).

Figure 4: Will global warming be more like water, soda, beer, … or vodka? Here we have four scenarios for future warming, assuming we just keep digging up fossil fuels, burning them, and leaving the released carbon dioxide in the atmosphere. In the first, the only feedback operating in the climate is the strongly negative “Planck” feedback (“hotter things give off more light”), which like a quenching glass of water simply satisfies your pang of thirst; just a little warming closes the imbalance between inflows and outflows of energy (blue line). In the second, the positive water vapor feedback (“hotter planets evaporate more water, which is itself a greenhouse gas”) and others makes warming more like soda, and you might end up drinking quite a Big Gulp to satisfy your thirst (red line). In the third, the deliciousness of the water vapor feedback gets amplified the more you warm, giving your drink a kick like a nice glass (or two, or three…) of beer (tan line). This state of affairs seems to describe the response of the CanESM5 model, which warms more than if its feedbacks stayed constant (green line). Finally, a climate with a higher feedback temperature dependence – a higher alcohol content? – can end up on quite a bender (gray line). Looking at past climates suggests that we are not at risk for this particular scenario. Figure credit: Jonah Bloch-Johnson, Maria Rugenstein, Martin Stolpe, Tim Rohrschneider, Yiyu Zheng, and Jonathan Gregory.

Is the addictive nature of the water vapor feedback important for global warming? Evidence from records of the past and computer models of the future suggest the Earth is likely not close to a runaway greenhouse. However, even if we are not at risk of “drinking” away our oceans, the road to warming may be paved with tipsiness. The green line in Fig. 4 shows the warming of the Canadian Earth System Model, Version 5 (CanESM5) if we continue to increase the carbon dioxide concentration until it reaches eight times the preindustrial level in 2250, which assumes that we continue to grow using mostly fossil fuels. Using the feedbacks and feedback temperature dependence of CanESM5, we can extrapolate warming out even further (tan line). The difference between the red and tan lines – between the soda and the beer – gives you additional warming that continues to increase with time.

While scientists have long appreciated the importance of understanding how “sweet” warming is for forecasting climate change in the coming century, it is becoming clear that after 2100, climate change will also depend on the booziness of warming. While 2100 once seemed like the unimaginable future, it will soon be within the lifetime of most of the newest humans. In the meantime, we keep spewing out CO2. In response, we may find that the Earth doesn’t just warm – it might also develop a taste for it.

#### Reference:

Bloch-Johnson J, Pierrehumbert RT, Abbot DS (2015) Feedback temperature dependence determines the risk of high warming. Geophysical Research Letters 42:4973-4980. https://doi.org/10.1002/2015EA000154

## How to analyse your forecast diary properly

As you are reading a science blog, I am sure you are interested in science, and either as a parent or during your childhood you will have seen these books aimed at children interested in the natural world. I’m thinking of colourful books with busy layouts and named along the lines of “100 Science Experiments For Little Explorers”, “365 Activities For Young Scientists” or “50 Cool Experiments For Rainy Days”.

A staple of these books is suggesting to start a weather diary as an activity. Depending on the targeted age groups, these might simply involve putting dashes next to a smily sun or a bubbly rain cloud, or be more challenging by asking to record more details and even write little reports.

Suggestions as to what to do with these records are much harder to find. This article suggests to make your weather diary even more interesting by recording not only what has actually happened but also what weather centres (such as the MetOffice) predicted to happen. Further, we will discuss how to do something useful with your records, namely subject them to a proper statistical analysis, thereby checking the quality of the forecasts in some sense.

The challenge is that your forecast diary will most likely not be amenable to the standard statistical tools that are taught in those moderately popular statistics courses for natural scientists. The data instances in your forecast diary will not be temporally independent, which renders the whole analysis a lot more interesting.

## The forecast diary

The diary we have in mind records not only the actual event (say “rain” or “no rain” today) but also the corresponding forecast from some weather centre of your choice, let’s say the Met Office. Rain is interesting because the Met Office issues probability of precipitation, or PoP forecasts. These represent the propability of seeing rain during a specific time interval (at a given location). The Met Office uses hourly intervals, but presumably you don’t want to enter records into your diary with such high frequency. As an alternative, you can consult the Dutch or the US Weather Service as they provide daily PoP’s but only for their respective countries. But at least the latter seems to be taking the maximum of the hourly PoP’s as a daily PoP (whatever the theoretical justification). This is something you could apply to the Met Office hourly PoP’s to get daily PoP’s for the UK.

The forecast is expressed in percent, rounded to the nearest multiple of ten. In fact, there are several forecasts available that apply to any given event, namely forecasts with different horizon, also called lead time. You may record forecasts for several lead times in your diary, a typical page of which might look like this:

Forecast Diary
Date Rain $L = 1$ $L = 2$ $L = 3$
01.01.20 Yes 20 10 10
02.01.20 No 10 10 10
03.01.20 No 10 0 5
04.01.20 Yes 50 40 60
05.01.20 No 20 10 10
06.01.20 Yes 60 60 70
07.01.20   50 40 50
08.01.20     10 0
09.01.20       10
09.01.20
:

This is how your forecast diary might look like on 6th of Jan, 2020. Here, Probability of Precipitation forecasts (in percent) up to a lead time of three days are recorded in separate columns with headings $L = 1$, $L = 2$ and $L = 3$.

## Reliability

Focussing on forecasts for a fixed lead time (say three days), a forecast diary can be used to check the reliability of the forecasting system. In the present situation, this means that if the forecast probability is $p$, then it should indeed rain with probability $p$. More precisely, considering all rows in the diary where the relevant forecast was $p$, we should see rain for a fraction $p$ of those rows; and this needs to be checked for all values of $p$ that can possibly occur. In our case, there are eleven values: $0, 0.1, 0.2, \ldots, 0.9, 1$.

In general, assume the forecasts take the values $\pi_1, \ldots, \pi_K$. Assume we recorded $N$ days in total, and for each $k = 1, \ldots, K$ let $R_k$ (or $N_k$, respectively) be the number of days where the forecast was equal to $\pi_k$ and there was rain (or no rain, respectively). Following our discussion, we expect that

$\frac{R_k}{ R_k + N_k} \cong \pi_k.$

You can plot the left hand side (the observed frequencies) vs the right hand side (the forecast probabilities) and check if you get something close to the diagonal; this is called a reliability diagram. Here’s an example of a reliability diagram, taken from Bröcker and Smith, 2007:

Rearranging the previous equation, we find

$(1 - \pi_k) R_k + \pi_k N_k \cong 0 \qquad \mbox{for } k = 1, \ldots, K. \qquad (*)$

Of course, if you compute the left hand side for your forecast diary, due to random fluctuations it won’t equal to zero even if the forecasting system was reliable.
Working out how large these deviations typically are and whether a given deviation is still acceptable is what happens in a statistical test, which is what we will discuss next. For more about reliability diagrams and graphical ways of representing the expected random fluctuations see Bröcker and Smith, 2007.

## The central limit theorem

We abbreviate the left hand side of Equation (*) as $Z_k, k = 1, \ldots, K$ and realise that if the rows in the diary were stochastically independent, we could apply the central limit theorem and obtain as an immediate consequence that the quantity

$t := \frac{1}{N}\sum_{k, j} Z_k Z_j C^{-1}_{k, j}$

would have a $\chi^2$-distribution with $K$ degrees of freedom (where $C$ is the covariance matrix of the $Z_1, \ldots, Z_K$). Note that $t$ represents the sum squared deviations we encounter in Equation (*) but in a weighted metric defined by the covariance matrix $C$.

The rows of the diary are not independent though. In fact the whole point of weather forecasting is that the weather today tells us something about the weather tomorrow. Fortunately, this does not preclude the use of the statistical methodology outlined above. The central limit theorem is still applicable provided the correlations between the relevant terms in the sum decay fast enough. That this is indeed the case is shown in Bröcker, 2018 (in a slightly different situation, but the basic idea carries over).

This decay is not due to any (assumed) decay in the temporal correlations of the weather due to chaotic dynamics but follows from the assumption that the forecasts are reliable. Remember, we want to know if the deviations in Equation (*) are large provided the forecasts are reliable; if they are, the assumption of
reliability is untenable.

It needs mentioning though that the covariance matrix $C$ is not what you would get if the rows in your diary were independent. This matrix has to be estimated from the data. How to do this is beyond this article but python code implementing the methodology can be found here. Look at the comments in rainscript.py and apply it to raindataL2.csv or, of course, your own forecast diary.

## References

Jochen Bröcker.
Assessing the reliability of ensemble forecasting systems under serial dependence.
Quarterly Journal of the Royal Meteorological Society, 144(717), 2018. https://doi.org/10.1002/qj.3379

Jochen Bröcker and Leonard A. Smith.
Increasing the reliability of reliability diagrams.
Weather and Forecasting, 22(3):651–661, June 2007.https://doi.org/10.1175/WAF993.1

## Our changing energy system: what happens at times of high demand?

It is getting near to Christmas, which means the decorations are coming out, the nights are drawing in, and even Scrooge has conceded that it’s probably time to put the heating on. In winter the UK has a significantly higher electricity demand than it does in summer, due to our increased need for heating and lighting. In past decades meeting this increased demand was relatively easy, as our generation was supplied from a more “traditional” fleet of coal, gas, and oil-fired power stations. These are relatively easy to control when they are on or off and can turn on very rapidly to meet a sudden peak in demand if the EastEnders Christmas special comes on. In order to meet the government’s proposed carbon reduction targets, the type of generators that are used to meet demand are rapidly changing. A transition is underway to more renewable sources such as wind, solar and hydro power. The operation of the renewable generation depends on the present weather conditions, as opposed to the TV scheduling (although perhaps online streaming services are helping the national grid by reducing TV viewership). It is therefore important to understand how the weather sensitivity of the current power system is likely to change when more renewable generation is introduced.

Recent research from the energy-meteorology research group has investigated various aspects of this problem using weather-dependent models of demand, wind power generation and solar power generation (available here if you’d like to have a play with the data). A problem with using measured demand and renewable generation data for this type of analysis is that there are large trends in the data due to the rapid expansion of renewable capacity, or due to changes in economic factors, human behaviour and the energy efficiency of every-day products. We therefore build reconstructions of a fixed power system configuration (i.e. taking the amount of installed renewable generation and level of demand in 2017) and then use a reanalysis dataset to reconstruct what the present day power system would have looked like with the last 40 years of weather. With these long datasets we can look at the meteorological drivers of power systems [1,2] and investigate thoroughly the changing weather conditions associated with extreme power system events as we put more renewables on the system [3,4,5]. Some examples of this are discussed below.

Figure 1: Anomaly composites of the mean of the ten most extreme peak loads from the 1980–2015 winter-mean. These are given for for 2 m temperature (first column) and 10 m wind speed (2nd column) for a UK power system with no renewables (NO-WIND) and one with 45GW of installed wind power. Mean-sea-level-pressure contours for the events are overlaid in black with a 4 hPa interval. The thick contour represents 1016 hPa. The ’H’ represents the location of the centre of the region of high pressure. Adapted from [3]

Figure 1 shows composites of the ten highest demand events experienced in the UK from 1980-2016 (using the synthetic time series discussed previously). We see that if there is no renewable generation installed on the system then peak demand events are associated with very cold near-surface temperatures and moderately low winds. The key ingredient here is the location of a high pressure system, with a strong pressure gradient over the UK bringing cold continental air from central Europe. If we were to look forward to planned future power systems including 45GW of wind power generation (around twice the amount currently installed) then the weather conditions causing largest system stress are likely to change. Figure 1 shows peak load events (i.e. the demand minus all available wind power generation) are still associated with high pressure, but there is a shift to high pressure events centered over the UK now being most important. These events lead to both a reduction in the magnitude of the temperature anomaly and an increase in the wind speed anomaly over the UK.

It is relatively easy to forecast the broad characteristics of energy demand, as we all tend to go to work and take holiday at certain times of the day/year. It is however challenging to forecast the type of events shown in Figure 1 at timescales greater than a few days. If information was available at timescales of 1 week to 1 month ahead (sub-seasonal timescales) then this would allow for planning decisions to be made about where power is going to be sourced from to meet high demand events [6]. Studies have shown that at these longer lead times, there is relatively more skill at forecasting large-scale weather conditions, rather than local anomalies at the surface. Because of this, some current work as part of the S2S4E project has investigated the potential for two types of large scale patterns to be used to provide information at sub-seasonal timescales for energy system resource planning.

The first pattern-based method is weather regimes. These are patterns defined based on large scale meteorological conditions (daily anomalies of 500hPa geopotential height). Each day during the extended winter season can be assigned into one of the patterns: the positive and negative phase of the North Atlantic Oscillation (NAO+, NAO-), Scandinavian Blocking (ScBL) and the Atlantic Ridge (AR). Rather than clustering on a large-scale field, the second method, newly developed for the project is called Targeted Circulation Types. This assigns each day into one of four patterns based on national energy data (using 28 countries across Europe). These patterns are: Blocked (BL), Zonal (ZL), a European High (EuHi) and the European Trough (EuTr).

Figure 2: Probability of each country’s demand being in the upper climatological tercile of demand in each weather regime (top) and Targeted Circulation Type (bottom). The TCT patterns used are constructed from normalised demand. See [1] for more details.

Each of these 8 weather patterns can be associated with a set of surface meteorological impacts (e.g., a set of demand, wind power or solar power anomalies). Figure 2 shows the probability of demand being in the upper tercile (top 30%) while each of the patterns has occurred (again using our reanalysis-based power system data to get a longer record for analysis). This shows that particular patterns are more likely to be associated to high demand events, especially those associated with European Blocking. Some patterns do not have a particularly strong link to European demand (e.g. the Atlantic Ridge). As well as this, the link to the impacted system is not as strong using weather regimes as using the Targeted Circulation Types. Current work is investigating if there is a tradeoff in the levels of predictability associated with the two different types of patterns. For further details see [1] and the S2S4E website.

References:

[1] Bloomfield, H. C., Brayshaw, D. J. and Charlton-Perez, A. (2019) Characterising the winter meteorological drivers of the European electricity system using Targeted Circulation Types. Meteorological Applications. (2019) https://doi.org/10.1002/met.1858

[2] van der Wiel, K., Bloomfield, H., Lee, R. W., Stoop, L., Blackport, R., Screen, J. and Selten, F. M. (2019) The influence of weather regimes on European renewable energy production and demand. Environmental Research Letters, 14. 094010. https://doi.org/10.1088/1748-9326/ab38d3

[3] Bloomfield, H., Brayshaw, D. J., Shaffrey, L., Coker, P. J. and Thornton, H. E. (2018) The changing sensitivity of power systems to meteorological drivers: a case study of Great Britain. Environmental Research Letters, 13 (5). 054028. https://doi.org/10.1088/1748-9326/aabff9

[4] Bloomfield, H. C., Brayshaw, D. J., Shaffrey, L. C., Coker, P. J. and Thornton, H. E. (2016) Quantifying the increasing sensitivity of power systems to climate variability. Environmental Research Letters, 11 (12). 124025. https://doi.org/10.1088/1748-9326/aabff9

[5] Drew, D. R., Coker, P. J., Bloomfield, H. C., Brayshaw, D. J., Barlow, J. F. and Richards, A. (2019) Sunny windy Sundays. Renewable Energy, 138. pp. 870-875. https://doi.org/10.1016/j.renene.2019.02.029

[6] White, C. J., Carlsen, H., Robertson, A. W., Klein, R. J., Lazo, J. K., Kumar, A., … & Bharwani, S. (2017). Potential applications of subseasonal‐to‐seasonal (S2S) predictions. Meteorological applications, 24(3), 315-325.  https://doi.org/10.1002/met.1654

## Climate change increases the “perfect storm” coastal flood potential

Some of the European low-lying coastal areas and river estuaries may see a future increase in flooding caused not only by sea-level rise but also by more frequent concurrent storm surge and heavy precipitation, we show in a new study published in Science Advances [1].

Figure 1: Compound flooding caused by the interaction between storm surge and heavy rainfall. A moderate storm surge can obstruct the gravity-feed drainage of pluvial or fluvial water into the sea. Or, rainfall on top of a storm surge inundated area can worsen the overall flooding. Image originally published on the BBC.

Coastal flood risk is usually calculated based on storm surge and extreme rainfall hazards separately. However, ignoring compound flooding arising from the local interaction between these two hazards may cause severe underestimation of the actual flooding risk in some areas.

The importance of considering compound flooding for planning coastal defences is shown, for example, by the past flooding management in Lymington [3], a port town in the southern UK. In December 1989, a storm surge led to widespread flooding in the town, causing damages to houses and the railway line. As a result, an extensive upgrade of the coastal defences took place; however, the potential for flooding resulting from the interaction of storm surge and pluvial flooding was not considered at that time. Ten years later in 1999, the Christmas Eve in Lymington was stormy: a cyclone not only caused a storm surge but, simultaneously, also very heavy rainfall leading to high discharge in the Lymington River. The new raised defences were activated and protected the town from the storm surge well; however, the prolonged storm surge did not allow the sea defences to be deactivated, thus trapping the water discharge of the Lymington River on the upstream side of the sea defences. This combination of events resulted in very deep flooding over the railway line and into residential and commercial development in Lymington.

Figure 2: Flooding around Lymington (UK) on December 24, 1999. Photo from The New Forest LIFE Projects.

Similar compound flooding happened, for example, in 2012 in The Netherlands [4], and in 2015 in Ravenna (Italy) [2] due to the cyclone Norbert causing concurring heavy rainfall and storm surge. In fact, coastal heavy rain often coincides with a storm surge because these two hazards frequently result from the passage of a cyclone. Thus, considering only one of the two hazards at a time, or assuming that they are independent, can result in a substantial underestimation of the flooding risk.

It is clear that in low-lying coastal areas and river estuaries, compound flooding should be considered to avoid underestimating the flooding risk. However, the need for considering the multiple flooding drivers simultaneously makes compound flooding risk assessment a rather complicated task. In addition, as climate changes, we need to understand how the compound flooding risk will change in the future, which is vital for adapting to flooding changes through – when necessary – building or upgrading flood defences. Thus, also future changes in sea level and precipitation need to be taken into account when assessing the flooding risk, making a compound flooding assessment extremely challenging.

In a new study, colleagues and I investigated, for the first time, how compound flooding may change in the future due to sea-level rise and changes in precipitation, storm surge, and their interaction.

Focussing on the European and North African coasts, we found that not only sea-level rise will cause an increase in the potential for compound flooding everywhere, but additional changes in compound flooding will be caused by changes in storms. In particular, the potential for compound flooding is projected to increase in northern Europe due to increasing precipitation intensities, mainly caused by a future warmer atmosphere that allows storms to carry more moisture.

The results of the study cannot be interpreted as a precise assessment of the local flooding risk, though it is evident that compound flooding will become more likely in the future if we do not take any action against climate change. The study aims at offering a large-scale view of the present and future potential for the compound flooding hazard. Identifying European regions potentially facing compound flooding in a warmer climate, it provides a basis for follow-up local compound flooding risk assessments and adaptation planning.

In locations prone to compound flooding, adapting to sea-level rise only might be insufficient, and additional changes in precipitation, storm surges, and their interaction should be considered to avoid too optimistic adaptation planning.

References

[1] Bevacqua, E., D. Maraun, M. I. Vousdoukas, E. Voukouvalas, M. Vrac, L. Mentaschi, and M. Widmann, 2019: Higher probability of compound flooding from precipitation and storm surge in Europe under anthropogenic climate change. Science Advances, 5, no. 9, eaaw5531, https://doi.org/10.1126/sciadv.aaw5531.

## The giant space plasma waves that can destroy our satellites

Everyday life is becoming more and more dependent on satellite services. From critical communications to forecasting and GPS, we would feel the impact of these lost services quickly. The location and accurate time provided by GPS is vital for navigation, the power grid, computers, phones, financial transactions… even the food we eat in the UK relies on timely transport.  (The BBC recently published an article on the impact of GPS which you can read here.) However, satellites are susceptible to several types of damage through space weather.

The region of space near Earth that is dominated by our planet’s magnetic field is known as the magnetosphere. This provides us and our satellites significant protection from the fast-moving and highly charged solar wind streaming out from the Sun. However, this barrier is highly dynamic as the solar wind constantly buffets us on its way through the solar system. This buffeting causes giant large-scale waves that can bounce around inside the magnetosphere, energising and transporting trapped electrons and posing a hazard to satellites that reside in the radiation belt region. The radiation belts are doughnut-shaped regions that contain many charged particles, trapped by the magnetic field and continually travelling around the Earth.

Figure 1: The Van Allen probes spent seven years observing the radiation belts, the region of energetic trapped particles that are hazardous to satellites. Credits: NASA Goddard’s Scientific Visualization Studio

Actually, there are multiple ways in which space weather can threaten satellite services. High energy particles such as “killer electrons” can penetrate and ionise individual components. Electrostatic charging and the sudden discharges can develop on internal or external surfaces, and communications through the ionosphere can be severely disrupted. However, only some of these are directly related to the giant plasma waves I study – ultra-low frequency or ULF waves. These waves affect the energisation and transport of high-energy electrons. Since satellite operators don’t like to advertise when they have spacecraft failures or the details of those failures, it’s difficult to pinpoint many occasions where these waves were the dominant factor. One well-known example is the failure of the two Telesat spacecraft in 1994, Anik-E1 and E2 [Lam et al, 2012]. We’re aware of this because of the sheer amount of disruption caused; these satellites carried virtually all Canadian television and a significant amount of communications capability. Energetic electrons accelerated by ULF waves damaged the satellites such that it took over six months to regain full services and a hundred thousand customers had to manually re-orient their satellite dishes towards the recovered spacecraft.

Figure 2: Perturbations at the magnetopause can drive waves that propagate inwards, disturbing Earth’s magnetic field. These ULF waves can reflect, bounce or form standing waves that can be measured at the ground. Credit: Sarah Bentley

Technically, we classify “ultra-low frequency” to be below 30 Hz, but I’m mostly interested in the 2-10 mHz range of waves. ULF waves are plasma waves, which means that as well as involving density compressions like a sound (or fluid) wave, they can also incorporate oscillations in the electric and magnetic field. All these oscillations are so large that we measure their period in minutes or hours and their wavelengths in thousands of kilometres – comparable to the radius of the Earth. These waves have their strongest effects when we see sustained wave activity. In this case, repeated electric field pulses can coincide with the motion of the multitude of electrons zipping around the Earth. The electric field exerts a force on the trapped electrons, accelerating them or moving them to different areas of the magnetosphere. So, predicting the occurrence, amplitude and extent of these, is an important aspect of forecasting the radiation belt environment. This would enable satellite operators to take steps to protect their spacecraft, for example by moving the satellite or shutting down vulnerable components.

Unfortunately, predicting the extent of these waves isn’t always that easy. Because they are so big, it’s more computationally feasible to simulate them numerically than other aspects of the magnetosphere. But it’s still very time-consuming, and we can’t run simulations that correspond to a given time in real life because we would need to know what the boundary of the magnetosphere looks like and what the driving solar wind is doing. Typically, we only have a single point measurement in the solar wind near the Earth, which is just not enough to fully describe an environment hundreds of thousands of kilometres wide. An easier approximation we have been making is to simply use an empirical, statistical model which gives median ULF wave power under different solar wind conditions [Bentley et al, 2019]. To our surprise, just using solar wind properties predicted the ULF wave power better than if we assume that we would see the same power from one hour to the next – our model was much more successful than anticipated! We found that the amount of energy seen in these waves changes with different solar wind properties to those we expected. This suggests that the driving of these waves is slightly more complicated than previously realised and that the processing of the magnetosphere may be more important.

Eventually, we expect that a model like this will improve radiation belt forecasting. This is becoming more important than ever – while the last decade or so has been particularly quiet for the radiation belts, there’s no guarantee this will last. In this time, we have become ever more dependent on satellites, and the way that we use them has made them more susceptible to radiation damage. Spacecraft now often use off-the-shelf rather than custom radiation hard components, and cheaper methods of getting into orbit means that they spend even longer in the radiation belts. So, we hope to understand and predict these giant waves better and discover more about the complex and weird behaviour in the area dominated by the Earth’s magnetic field.

## The latest on aerosol radiative forcing

Aerosols are tiny liquid or solid particles suspended in the Earth’s atmosphere. Some aerosols form naturally, like the sea spray emitted by breaking waves, the mineral dust that form sandstorms, or smoke from wildfires. But human activities, from combustion of fossil fuels, cement manufacturing, fertilisers, and agricultural and forest clearing fires also emit aerosols into the atmosphere.

Once in the atmosphere, aerosols affect the energy budget of the Earth by reflecting and absorbing sunlight. They also play an important role in the formation of liquid clouds, and aerosols from human activities lead to the formation of more reflective clouds, which may also have a different liquid water content or be slower to rain out. The extent to which aerosols from human activities modify the energy budget of the Earth is measured by a quantity called aerosol radiative forcing.

Scientists have studied aerosols for a long time. Their role in the formation of fog was identified by the Scottish meteorologist John Aitken in 1880. In the 1960s and 1970s, scientists observed ships making clouds thicker (the so-called ship tracks) and paper mills affecting precipitation downwind of their location. But the global extent of aerosol perturbations was only revealed during the 1980s and 1990s when ground-based observations networks, aircraft campaigns, and satellite instruments tremendously increased our ability to observe aerosol properties.

Figure 1: Distributions of aerosol optical depth, a measure of aerosol loading of the atmosphere, as estimated in (left) 1984 based on knowledge of the time; and (right) 2019 based on a numerical model of atmospheric composition that ingests satellite observations of aerosols.

Figure 1 shows how understanding of global aerosol patterns has dramatically improved thanks to better observations and better models of aerosol sources, transport, and removal. The research done over the past 40 years identified aerosol sources and their variability, characterised the wide spectrum of aerosol properties as they “age” in the atmosphere, and discovered the complexity of the natural atmosphere.

Figure 2: Uncertainty ranges for aerosol radiative forcing from interactions with radiation (ari, left in blue columns) and interactions with clouds (aci, right in orange columns) in past Assessment Reports (AR) of the Intergovernmental Panel on Climate Change. The first report did not quantify aerosol radiative forcing but noted that it was potentially substantial.

That improved understanding quickly translated into the first estimates of aerosol radiative forcing, but progress then seemed to stall. Figure 2 shows estimates of the uncertainty ranges of aerosol radiative forcing as assessed by successive reports of the Intergovernmental Panel on Climate Change. Despite the large, international intellectual effort dedicated to observing, understanding, and modelling the impact of aerosols on climate, uncertainty in aerosol radiative forcing has not changed much over the past decades.

Over the past year, I have led more than 30 colleagues, each bringing complementary expertise in the many ways we observe and model aerosols, to review the scientific literature on aerosol radiative forcing. Our aim was to take a fresh and comprehensive look at present understanding of aerosol radiative forcing and identify prospects for progress on some of the most pressing open questions. On 1 November 2019, the scientific journal Reviews of Geophysics published a preliminary version of our review article .

We found that important aerosol radiative forcing mechanisms are getting better constrained. The balance between aerosol reflection and absorption of sunlight, and the degree to which aerosols from human activities increase the number of liquid cloud droplets, are now well understood. In addition, a series of recent observations, including from volcanic eruptions, found that cloud liquid water content was less sensitive to aerosol perturbations than previously thought.

But important gaps in our understanding remain and those gaps explain why uncertainty in aerosol radiative forcing remains large. Some of the most important research questions are:

• What were the aerosols levels before industrialisation dramatically increased human emissions? How polluted was the natural atmosphere by wildfires and emissions from plants, deserts, and oceans? Those questions are difficult because preindustrial aerosols have not been observed, and numerical models incompletely represent the sources and properties of natural aerosols.
• Do aerosol perturbations increase cloud fraction, as predicted by large-scale models of the atmosphere? Or is the picture more nuanced, as suggested by satellite observations or predicted by smaller-scale, more detailed models? Those questions are difficult because cloud fraction is a poorly-defined quantity and depends on the scale at which clouds are observed and simulated.
• Do ice clouds respond to perturbations of aerosol amounts by human emissions? And if so, what is the associated radiative forcing? Those questions are difficult because ice crystal number is poorly observed, and aerosol-ice interactions are not well understood theoretically.

To quantify the remaining uncertainties, my colleagues and I worked out that aerosol radiative forcing offset at least a fifth and up to half of the radiative forcing by greenhouse gases, with 2 chances out of 3 that the right answer lies between those two bounds. We also identified new approaches that hold great promise to make further progress:

• Statistical methods to compare multiple model configurations to better, more numerous observations improves our understanding of the sources of uncertainty in models.
• Increases in computing power allow simulations of the whole atmosphere with detailed models of aerosol-cloud interactions, which better represent subtle processes.
• Natural laboratories like volcanic eruptions and ship tracks provide opportunities to understand aerosol perturbations of ice clouds.
• Analyses of sediments to find traces of ancient charcoal give unprecedented insights into past fire activity.
• Observed changes in surface temperature, ocean heat content, or sunlight levels at the surface can be used to constrain aerosol radiative forcing. The use of differences between the two hemispheres, or between time periods where aerosol pollution increases or decreases, may make those inferences stronger.

Attempting to quantify aerosol radiative forcing has given us fascinating insights into the atmosphere of our planet and its climate system. Many questions remain wide open and provide much to excite the curiosity of the best physicists, chemists, and mathematicians. If a new group of aerosol scientists repeats our review in 10 years’ time, I hope they will be able to marvel at the progress accomplished.

References:

Bellouin, N., Quaas, J., Gryspeerdt, E., Kinne, S., Stier, P., Watson‐Parris, D., et al (2019). Bounding global aerosol radiative forcing of climate change. Reviews of Geophysics, 57. https://doi.org/10.1029/2019RG000660

Tanre, D., J.-F. Geleyn, and J. Slingo (1984), First results of the introduction of an advanced aerosol-radiation interaction in the ECMWF low resolution global model, in Aerosols and Their Climatic Effects, edited by H. Gerber and A. Deepak, pp. 133–177, A. Deepak, Hampton,Va.

Inness, A., Ades, M., Agustí-Panareda, A., Barré, J., Benedictow, A., Blechschmidt, A.-M., Dominguez, J. J., Engelen, R., Eskes, H., Flemming, J., Huijnen, V., Jones, L., Kipling, Z., Massart, S., Parrington, M., Peuch, V.-H., Razinger, M., Remy, S., Schulz, M., and Suttie, M.: The CAMS reanalysis of atmospheric composition, Atmos. Chem. Phys., 19, 3515–3556, https://doi.org/10.5194/acp-19-3515-2019, 2019.

## Howling Space Gales and why we should photograph them.

By: Luke Barnard

Most people are familiar with the fact the Sun emits a range of electromagnetic radiation (e.g. sunlight), and that this radiation is necessary to sustain life on Earth as we know it. What is less well known is that alongside the Sun’s electromagnetic radiation, it also generates a wind of plasma that continuously blows out through the Solar System, with speeds of 250 km/s to 750 km/s.

This solar wind impacts our everyday lives through its effects on the technology we increasingly depend on; particularly spacecraft in orbit around Earth. We rely on satellites for critical services such as communications, GPS, and weather forecasting. When services like these are disrupted, it can have both expensive and dangerous consequences [1].

During periods of intense solar wind activity, it squeezes and shakes Earth’s magnetic field. This produces energetic charged particles which are harmful to satellite electronics and can also make it difficult to maintain radio communications with them. Depending on how intense the solar wind is, satellites can be temporarily or permanently damaged, with knock on impacts to the services they provide.

Space Weather Forecasting grew out of the need to understand and predict when situations like this would occur. A key challenge in space weather forecasting is to be able to forecast the solar wind flow throughout the Solar System. This is difficult because there are only a handful of spacecraft able to measure the solar wind, and these only measure it at single points which are vastly separated. By way of analogy, it is like trying to forecast the weather at Reading, with only weather observations at a few other far away cities, like Exeter, Manchester and Brighton. The limited information is still useful, but there is a lot that can happen in between and a lot of uncertainty.

Our research [2] aims to help solve this problem by using images of the solar wind plasma to characterise the solar wind flow near the Sun. This would be an extra source of information on the solar wind flow, which we could use to help improve computer models that forecast the solar wind.

Figure 1: This shows the relative locations of Earth, STEREO-A and STEREO-B. The purple shaded regions show the field-of-view of the inner Heliospheric Imager camera on STEREO-A.

NASA’s STEREO mission consists of two spacecraft which are in Earth-like orbits, but they drift relative to Earth [Figure 1], so that they can observe the space between the Sun and Earth. On each spacecraft are a pair of cameras called the Heliospheric Imagers, which produce images of the solar wind plasma [Figure 2]. The cameras record visible sunlight that has scattered off of electrons in the solar wind. Interpreting the images is tricky because there are electrons and visible light everywhere in space, and so we don’t actually produce an image of a specific feature or object. But, because we understand the physics of sunlight well, and of how sunlight scatters off of electrons, we are able to use these images to identify regions where there are relatively more electrons, and a denser solar wind.

Figure 2: A movie of heliospheric imager images from July 2008. Movie obtained from the UK Solar System Data Centre

Our aim was to show that variability in the images could be statistically related to the direct single point measurements of solar wind flow observed by other spacecraft. This would be the first step in creating and calibrating a technique to estimate the solar wind flow directly from the images.

We compared the solar wind point measurements and images directly, computing the correlation between variability in the images recorded by STEREO-A with the solar wind speed measured directly at Earth, STEREO-A, and STEREO-B. We found that there is a strong correlation between variability in the images and the solar wind speed observations at the three spacecraft, but that the correlation was largest when a delay was applied between the image and solar wind observations. This delay was different for each pair of spacecraft, and changed in time in a way that can only be explained by the orbits of the spacecraft. Based on this statistical analysis we have concluded that we probably can trace the flow of the solar wind in the Heliospheric Imager data. Our next step is to investigate how to best compute a reliable estimate of the solar wind speed directly from the images.

## Trading Evil lasers for MAGIC Doppler lidars

By: Janet Barlow

Lasers may have an evil reputation in Hollywood, but they are very good for observing urban meteorology. We recently took part in the MAGIC project field campaign in London, deploying a Doppler lidar to measure wind-speed around tall buildings.

Just like a duck in water, a tall building causes a wake behind it. The wake can be 100s of metres long downstream, causing reduced wind speeds and increased turbulence. Wakes can thus affect air quality, so it is important to represent them in pollutant dispersion models.

Recently we reported on wind tunnel experiments where we measured flow around a model tall building at the MAGIC project experimental site. One question was whether the wake affected natural ventilation of the test building at the centre of the site. Measuring flow around actual tall buildings is impossible using traditional meteorological instruments like cup anemometers: they are simply too small to measure the whole wake. Instead, we used a Doppler lidar which can measure wind-speed remotely over a wide area (Drew et al. 2011).

Figure 1: Principle of infra-red Doppler lidar operation. Image taken from https://www.hko.gov.hk/publica/wxonwings/wow018/wow18e.htm

The principle behind radar observations of rainfall used for a weather forecast is that a pulse of electromagnetic radiation of a certain wavelength is beamed out into the atmosphere (Figure 1). A lidar uses infra-red light that interacts with particles of a similar size to the light wavelength. Some light is scattered back to the instrument and measured. But the backscattered waves are shifted in frequency by an amount proportional to the wind-speed blowing the particles around. This is the same “Doppler effect” that we hear when an ambulance goes by and its siren seems to change pitch: the soundwaves change wavelength in proportion to its speed. One advantage of using infra-red frequencies is that lidars are eyesafe. Not evil at all!

Figure 2: Photo showing MAGIC experimental site. The tall building (height: 81 m) and the lidar (white box) are highlighted with a red circle. The London Eye is on the far left and the Shard is on the far right.

We placed our Doppler lidar on the roof of a building at the MAGIC experimental site in London (Figure 2). At a height of 27 m we had a good view above most rooftops. We scanned the laser beam horizontally in a circle, meaning that laser light was reflected from tall buildings, allowing us to locate them.

Figure 3: Lidar horizontal scan of local wind-speeds minus the average wind-speed across the whole scan. The wind direction was north-westerly. The building is shown as a red square and its wake is the yellow area to the south-east of it.

Figure 3 shows a horizontal scan of wind-speed measured by the lidar. The velocity measurement at each pixel has been subtracted from the average velocity across the whole scan (NB: as velocity is negative towards the lidar, a wake appears as a positive difference). The wake is approximately 150 m long, which means the test building is definitely affected by the wake – it is 85 m away from the tall building. Flow around it is weaker and more turbulent, affecting pollutant levels and the ability to ventilate rooms through open windows.

So, does a wake measured around a real building resemble wind tunnel measurements? We also found that the building wake in the wind tunnel was long enough to affect the test building – but how much was wind-speed reduced, compared to if the tall building was not there? The wind tunnel experiments suggested around 40% reduction at the location of the MAGIC test building (Hertwig et al. 2019); the lidar measurements for our case study suggest around 25%. With 6 months of data, we have many more cases to analyse to quantify wake behaviour under different weather conditions.

This amazing instrument allows us to “see” urban winds and provides invaluable data to improve forecasting and building design. But we definitely don’t need to attach our lidar to a shark’s head. That would just be evil.

Thanks to Eric Mathieu, Elsa Aristodemou, Jess Brown, Ian Read and Selena Zito for technical assistance.

## Don’t (always) blame the weather forecaster

There are (I am sure) numerous metaphors that suggest that a small, almost immeasurable event, can have a catastrophic outcome – that adding the proverbial straw to the load of the camel will break its back. In 1972, the mathematical meteorologist Ed Lorenz famously gave the presentation, “Does the flap of a butterfly’s wings in Brazil set off a tornado in Texas?” Unlike for folks who do keep a domestic camel, this title was not intended to be interpreted literally, but instead to ask how a system like the Earth’s atmosphere is affected by vanishingly small perturbations. But is it possible for a butterfly’s flap to really have consequences? Without the ability to experiment on two or more otherwise identical Earths, demonstrating this is impossible.

Learning from computer simulations

Atmospheric scientists are acutely aware that computer-derived forecasts are sensitive to the ‘initial conditions’ provided to them. Modern weather forecasting is done by representing the atmosphere at an initial time with vast sets of numbers stored inside a computer (this set is called the initial conditions of the model). The computer marches this state forward in steps into the computer’s version of the future. The rules that the computer uses to do this task boil down to Newton’s laws of motion (i.e. how forces acting on air masses change their motion), and other processes that affect the behaviour of the atmosphere, like heating and cooling by radiation and by condensation/evaporation of water. Unlike in the real world, it is possible in the computer to create two identical sets of initial conditions apart from small differences, and then to let the computer calculate the two possible future states.

Sensitivity to initial conditions

So, what do scientists find from these experiments? At first the forecasts are virtually indistinguishable, but at some time they start to show noticeable differences. These appear typically on small scales and then start to affect larger scales (known as the inverse energy cascade). Lorenz discovered this serendipitously in the 1950s when he ran simplified weather simulations on a research computer (a valve/diode-based Royal-McBee LGP-30 with the equivalent of 16 kilobytes of memory). He found that if he stopped the simulation, and restarted it with similar, but rounded, sets of numbers representing the weather, the computer simulated weather patterns that became very different from those that are forecast had he not stopped and restarted the simulation. Lorenz had discovered sensitive dependence to initial conditions (or colloquially, the “butterfly effect” in connection with the title of his presentation). Faced with two such different outcomes, which one, if any, is the better forecast? Hmm …

Figure 1:

Numerical solutions of two x, y, z trajectories obeying the (non-linear) Lorenz-63 equations to demonstrate sensitive dependence to initial conditions (red and yellow lines/points). At t = 0 the initial conditions are indistinguishably close and at t = 3 the two trajectories virtually overlap. At t = 6 small differences appear, which become more obvious at t = 9. By t = 12 and t = 15 the two trajectories are so different that they occupy separate branches. The beauty of the structure that emerges by solving the Lorenz-63 equations is quite amazing. For the record, the Lorenz-63 equations are: dx/dt = σ(yx), dy/dt = –xz + rx y, and dz/dt = xy bz, with σ = 10, r = 28, and b = 8/3. The multiplication of one of x, y, z with another such variable gives these equations their non-linear property.

Try this at home

This effect is also seen in simple non-linear equations. In 1963, Lorenz published a seminal work, “Deterministic non-periodic flow”, where he introduced some equations that describe how variables, x, y, and z change in time. These equations may be regarded as representing a highly simplified version of the atmosphere. It is only possible to solve these equations approximately with the help of a computer (note to reader – try this, it’s fun!). One can visualise the solution by taking particular x, y, and z values at a given time as the co-ordinates of a point in space. Joining the points up in time shows the forecasts as trajectories, and one may think of different positions as representing different kinds of weather. Figure 1 shows two such trajectories (red and yellow), whose initial conditions are nearly identical at time t = 0. As time progresses, they diverge, slowly at first, until by t = 12 they represent completely different states (note the resemblance to a butterfly).

Ensemble weather prediction

Scientists routinely run large models from many initial conditions, each subject to a slight variation – a technique called ensemble forecasting. The initial conditions differ by amounts believed to be around the level of uncertainty that the weather is known using observations and previous forecasts. These are combined in a physically consistent way, using data assimilation (which is my area of research). As a rule of thumb, differences seen in the small-scale weather forecast patterns emerge first. Indeed if the forecast grid is small enough to resolve cloud systems then the ensemble members will likely first disagree in the forecast of convective events, like showers and thunderstorms. This is why patterns of convective precipitation are so hard to predict beyond a few hours. One forecast may predict heavy rain at a particular location between 4.00 and 4.10pm, another between 4.30 and 4.35pm, and another may predict no heavy rain. Ensemble forecasting allows forecasters to understand the range of likely outcomes (usually all ensemble members will predict heavy showers, but with slightly different locations), and to give probabilistic forecasts for individual locations. While small-scale features will differ, large-scale weather patterns (such as high and low pressure systems) are usually predicted accurately at these early stages. As forecast time progresses the uncertainty develops in larger scales and eventually the forecast of the large-scale systems become unpredictable.

Fundamental limits

As a rule of thumb, km-scale motion is predictable to no more than about half-hour, 10 km-scales to about one hour, 100 km-scales to about 10 hours, 1000 km-scales to about one day, 10000 km-scales to about four or five days, and the largest scales no more than about a week or two. In extra-tropical regions, for example, there is a particular kind of atmospheric instability (baroclinic instability) between scales of around one to three thousand km which can lead to a lowering of predictability on those scales, although observing the weather at these scales is given special attention so that the uncertainty at these scales is reduced in the initial conditions.

[We should note that climate models make projections many years, decades, or centuries into the future and use the same building blocks as weather models. Climate models though predict different things: long-time averaged conditions rather than the weather at particular times, which is thought to be very useful as long as realistic forcings (e.g. the radiative forcing associated with changes in greenhouse gas concentrations in the atmosphere) are known.]

Room for improvement?

So what hope is there of improving weather prediction given these fundamental limits? There are other factors that can be improved. The spread in the ensemble’s initial conditions can be reduced with more observations and better assimilation. Model error can also be reduced. No model is perfect, but there is room for improvement by decreasing the grid size and time step (severely restricted by cost and available computer power), and by improving the representation of physical processes (also restricted by computing and on research activity).  While scientific and technological barriers can be broken, the fundamental limits of nature cannot. As the air motion of the butterfly’s flap mixes with all the other fluctuations, it is impossible to say exactly how it will change the course of the atmosphere, just that it will.

## High-resolution insights into future European winters

Figure 1: Observed UK rainfall anomaly as a percentage of 1981-2010 monthly average for (a) December 2013, (b) January 2014, and (c) February 2014. Figure from Huntingford et al. (2014).

Most – roughly 70% – of Europe’s winter rainfall is brought by extratropical storms, which are steered our way by the westerly North Atlantic jet stream. The wettest winters, such as 2013/14 (Figure 1), are often when Europe is at the receiving end of a veritable convoy of storm systems, the human and economic impacts of which are felt far and wide.

An analogy: the sharpness of a digital photograph depends on how many pixels comprise it – in other words, on the camera’s resolution. The more pixels; the higher the resolution; the sharper the image. Climate models break down Earth’s atmosphere into three-dimensional pixels called grid cells. With a relatively low number of large grid cells (each typically 100-200 km wide), simulated weather systems not only appear pixelated when visualised, but are actually not all that realistic; their flows of air, heat and moisture don’t properly resemble those in the real world. This limits our confidence in using these models to make predictions. However, developments in high-performance computing have enabled the size of a climate model’s grid cells to be shrunk (to roughly 25 km in our case), thereby increasing their number and enabling the simulation of air flows over mountainous terrain, weather processes, and other aspects of atmospheric variability in more detail.

In a recent paper published in Journal of Climate, we address two questions. Does increasing a model’s atmospheric resolution improve the fidelity of simulated European winter hydroclimate? How do high-resolution future projections differ, if at all, from those at low-resolution? We compared simulations with low- and high-resolution versions of the same climate model – Met Office’s Hadley Centre Global Environmental Model (version 3) Global Atmosphere 3.0 (hereafter HadGEM3-GA3; Walters et al. 2011) – to establish exactly what the impact of resolution is on the North Atlantic jet and on downstream storm activity and precipitation. At the lowest resolution (‘N96’), the latitude-longitude grid is made up of grid cells each 135 km. At mid- (‘N216’) and high-resolution (‘N512’), grid cells are each 60 and 25 km, respectively. We’ll focus here on how the North Atlantic jet behaves at different model resolutions.

Figure 2: Frequency of North Atlantic eddy-driven jet latitude in reanalyses and HadGEM3-GA3 under historical climate (upper panel) and the projected future change (lower panel). We use ‘N’ notation to describe resolutions: ‘N96’ (135 km), ‘N216’ (60 km) and ‘N512’ (25 km). Figure adapted from Baker et al. (2019).

What about future projections? Under climate change (RCP 8.5), at all model resolutions, southern jet occurrences decrease but northern jet occurrences increase (Figure 2, lower panel). The upshot of this is fewer storms making landfall over southern Europe and more across northern Europe towards the end of the twenty-first century. These climate change consequences are significantly enhanced by increased resolution. Crucially, this reveals the extent to which lower-resolution models may have previously underestimated aspects of the jet’s response to climate change, and thereby changes in winter storms and precipitation, and their associated hazards. There is much more work to do: further studies investigating other models and climate change scenarios are needed, but our study offers insight into how high-resolution might bring the picture of Europe’s future winters into sharper focus.